Both log LLM calls, track cost and version prompts, and both offer a gateway. Helicone has been in maintenance mode since March 2026 and new cloud sign-ups are disabled; Braintrust is independent, raised an $80M Series B in February 2026 and is still adding evaluation features. For most teams the question is when to move off Helicone, not which to adopt.[1][6]
Braintrust is an active, hosted evaluation platform built around experiments and online scoring; Helicone is an open-source LLM request logger and AI gateway that has been in maintenance mode since March 2026, with new cloud sign-ups disabled. On 19 October 2026 Helicone said it is no longer part of Mintlify, with no new owner named. For a new deployment Braintrust is usually the practical choice; existing Helicone users should plan for the gateway and logging they rely on.
| Decision factor | Braintrust | Helicone |
|---|---|---|
| Ownership | Independent; $80M Series B in February 2026 | Acquired by Mintlify in March 2026; on 19 October 2026 said it is no longer part of Mintlify, no new owner named |
| Status | Active; several new features in public preview | Maintenance mode since March 2026; no feature changelog entries after November 2025 |
| Evaluation | Experiments, online scoring, LLM and code scorers, human review | Not an evaluation framework, by its own docs; Python evaluators turned off October 2026 |
| Gateway | Braintrust Gateway, public preview, free, no SLA | AI Gateway for 100+ providers with failover, caching and rate limits |
| Self-hosting | Enterprise only, data plane only; Braintrust runs the control plane | Apache-2.0; Docker, Compose or Helm |
| Agent tracing | One trace per run, including tool calls and handoffs | Header-based session grouping; no span model for agent steps |
| Free tier | Starter: 1 GB processed data, 10K scores, 14-day retention | Hobby: 10K requests, 7-day retention, existing accounts only |
| First paid tier | Pro: $249 a month | Pro: $79 a month, existing accounts only |
| Capability | Braintrust | Helicone | Prefactor |
|---|---|---|---|
| Observe | |||
| LLM call tracing | ✓ | ✓ | ✓ |
| Multi-step agent tracing (tool calls, nested spans) | ✓ | Partial | ✓ |
| Cost tracking | ✓ | ✓ | ✓ |
| OpenTelemetry support | ✓ | Partial | ✓ |
| Evaluate | |||
| Offline evals (datasets, experiments) | ✓ | Partial | ✓ |
| Online evals on production traffic | ✓ | Partial | ✓ |
| LLM-as-a-judge scoring | ✓ | Partial | ✓ |
| Code-based evaluators | ✓ | — | ✓ |
| Human review and annotation | ✓ | Partial | ✓ |
| Prompt versioning and playground | ✓ | ✓ | — |
| Act on findings | |||
| Runtime controls on live agent actions | — | Partial | ✓ |
| Gateway or proxy in the request path | ✓ | ✓ | — |
| Licence and deployment | |||
| Open source | Partial | ✓ | Partial |
| Self-hosting | Partial | ✓ | Partial |
Mintlify acquired Helicone in March 2026 and moved it to a maintenance-only phase covering patches, bug fixes and new model support, with an offer to help customers move to other platforms. Cloud sign-ups have been disabled since 3 March 2026. In May 2026 an AWS account lock took Helicone's dashboard and log processing down for about four days while the proxy stayed up. On 19 October 2026 the site began saying Helicone is no longer part of Mintlify, with details to come.
Braintrust is independent and shipping: Patterns, Topics, a thread view and workflow evals arrived between September and October 2026, several of them in public preview. A new team cannot open a Helicone cloud account, so for a new deployment the comparison is mostly academic.
Helicone sits in the request path. Its proxy and AI Gateway give one OpenAI-compatible endpoint for more than 100 providers, with failover, caching, rate limits and per-request cost from a model registry. Its prompt versioning deploys prompts by ID through the gateway. Its documentation says it does not run evaluations; scores from external tools can be posted to requests.
Braintrust is built around evaluation. You define a dataset, a task and scorers, run experiments and compare, then apply the same scorers to sampled production logs. Human review supports multiple and blind reviewers. Loop, an in-product agent, reads results, writes scorers and runs evals. Braintrust's own gateway is in public preview with no SLA, and its earlier AI proxy is deprecated.
Agent tracing differs in depth. Braintrust records each agent run as one trace, including tool calls, handoffs and nested model calls, with integrations for the OpenAI Agents SDK, Claude Agent SDK, LangGraph and others. Helicone groups requests into sessions by a header path, without a span model for agent steps.
Cost tracking overlaps. Helicone computes exact per-request cost through its gateway's model registry, prices more than 300 models and alerts on cost thresholds. Braintrust captures token usage and cost per traced call, alerts on cost, and in October 2026 added a coding-agent insights dashboard in public preview.
Since September 2026 Braintrust's human review also supports blind reviews and required scores, and application logs sit alongside traces.
Helicone's async logging is built on OpenLLMetry. Braintrust accepts OTLP traces and lists OpenTelemetry and Traceloop OpenLLMetry among its integrations, so OpenTelemetry-based instrumentation can be repointed rather than rewritten. Teams on Helicone's proxy integration will need to add tracing through Braintrust's SDK wrappers or framework integrations instead.
Plan the gateway separately. Braintrust's Gateway covers OpenAI, Anthropic, Google, AWS and others with caching and logging, but it is a preview with no SLA. If routing and failover in production depend on Helicone today, decide whether a preview gateway is acceptable before you switch. Prompts versioned in Helicone need to be moved to Braintrust's prompt versioning, which can be called from code or the REST API.
Self-hosting is the other difference to weigh. Helicone self-hosts fully under Apache-2.0. Braintrust self-hosting is Enterprise only and covers the data plane; Braintrust always operates the control plane, including the UI and authentication.
As of 20 October 2026, Braintrust's Starter plan is free with 1 GB of processed data, 10K scores and 14-day retention, and Pro is $249 a month with 5 GB, 50K scores and 30-day retention. Every plan includes unlimited users. Helicone lists Hobby free with 10K requests, Pro at $79 a month and Team at $799 a month, but those apply only to existing accounts while sign-ups are disabled. See Braintrust's pricing and Helicone's pricing.
Beyond the included amounts, Braintrust charges $4 per extra GB on Starter and $3 on Pro, and $2.50 or $1.50 per 1,000 extra scores. Pro also includes $100 a month in model credits.
We sell the layer this section describes. Read it with that in mind.
Stated plainly, since Prefactor publishes this page: Prefactor is an agent evaluation platform that scores each step of a run and acts on what it finds through runtime policies on tool calls and data access, approval holds for risky actions, and a kill switch for a run, an agent or a team. Braintrust scores after logging and does not block live actions; Helicone's gateway can block prompt injection for OpenAI models only. Neither holds a tool action for approval. Prefactor has no prompt playground or prompt registry, no gateway, and its platform is hosted, so it does not replace Helicone's routing or Braintrust's prompt iteration. See the full comparison of 15 platforms.
Reviewed against public product and documentation pages on October 20, 2026. If a vendor has changed a feature, product name, or positioning since then, send a correction and we will update it. Numbered source links in the page body point to the ordered sources below.
Prefactor scores each step of every agent run and can hold or stop an action when a score says it should not happen. Start on the free tier: 5,000 spans a month, no credit card required.
Prefactor helps teams observe, evaluate, and improve their AI agents in production, across every framework and provider.